Enrolling in a selective college STEM program pays off more for academically marginal students – even though they are less likely to graduate, Cornell economics research finds.

The analysis found that students who weren’t as ready for science, technology, engineering and math coursework were up to 18 percentage points less likely to complete STEM degrees than better-prepared peers at a flagship public university in Colombia. But their average future earnings – including graduates and dropouts – increased by as much as 40% compared to similarly qualified applicants rejected by the programs. For better-prepared students, the added earnings benefit from enrolling was minimal.

The findings lend support to policies that seek to broaden access to selective STEM programs instead of prioritizing graduation rates, suggesting they could help reduce income inequality.

“If colleges want to maximize their graduation rate in their STEM programs, then they would admit the most-prepared students,” said Evan Riehl, assistant professor in the Department of Economics and the ILR School. “But if their goal is to benefit society – meaning to admit students who are going to have the highest average benefit from enrolling in STEM programs – then our findings suggest that these colleges are actually better off admitting students from disadvantaged backgrounds even if they have relatively less preparation.”

Riehl’s paper, “The Returns to STEM Programs for Less-Prepared Students,” published in the May issue of the American Economic Journal: Economic Policy, is co-authored with Kevin Ng, Ph.D. ’22, a research analyst at CNA, a nonprofit research and analysis firm.

Research has shown STEM degrees generate higher earnings than other fields and boost innovation, driving interest in growing the number of STEM students around the world. Selective STEM programs, however, are known for weeding out underperforming students and for high dropout rates, prompting debate about potential mismatches for less-prepared applicants, who on average are more socioeconomically disadvantaged.

“A growing body of research finds that students from disadvantaged backgrounds tend to benefit from going to more selective schools,” Riehl said. “But some believe if there is a mismatch, it’s most likely to be in STEM majors, where research shows academic preparation matters the most.”

Investigating the benefits of enrolling in such programs, regardless of the graduation outcome, Riehl and Ng obtained data on applicants to all programs at Universidad del Valle, or “Univalle,” in Cali, Colombia, from 1999 to 2004. They linked that information to administrative records providing students’ scores on entrance exams across subjects, enrollment and graduation outcomes, and average monthly earnings roughly 15 years later, in 2017.

The researchers first compared outcomes for students with similar levels of academic preparation and likelihoods of completing STEM degrees based on test scores – the sole basis for highly competitive admissions at Univalle. They also compared outcomes for students who were either barely admitted or barely rejected by Univalle, including during a period when the university doubled cohorts in certain STEM subjects, thus enrolling more students with lower levels of preparation.

The analysis confirmed academic preparation’s importance, consistent with prior research. Only about one-third of the marginal students who enrolled in Univalle’s engineering and natural science programs graduated – a rate comparable to those at flagship U.S. universities, the authors said. Less-prepared students were 9 to 18 percentage points less likely to complete STEM degrees than their more-prepared peers.

But surprisingly, Riehl said, lower completion rates did not translate to lower average earnings more than a decade later. Compared to similar marginal applicants who were not accepted to a Univalle STEM program, those who enrolled earned 30% to 40% more. Among better-prepared students, the earnings return from enrolling was close to zero.

That gap may be explained, Riehl and Ng propose, by the students’ “counterfactual” schooling options – where they enrolled if rejected by Univalle. Research on graduation rates often assumes that students rejected by a selective STEM program will pursue a STEM degree elsewhere, Riehl said. But for those who barely missed out on admission to Univalle, the data showed that their fallback was more likely to be a lower-paying major or a technical school, not another college STEM program.

An important caveat, the researchers said, was that the significant earnings gains for less-prepared students might be concentrated among those who had managed to graduate. Policies striving to improve STEM skills at younger ages could help ensure that returns are spread more evenly, they said. But Riehl and Ng concluded that focusing only on graduation rates can lead to an overly pessimistic view of selective STEM programs’ potential benefits.

“STEM programs can play an important role in reducing earnings inequality,” they wrote, “among students who arrive at college with different levels of academic preparation.”